Jul 12, 2023 4 min read

I'm Giving Away My Must-Try ChatGPT Prompts

By now, everyone in the media business is using generative AI for their work in some form or another.

So much so that last week, Make Lemonade—formerly Niche Website Builders—announced it had become insolvent due to cashflow problems and had gone into administration.

The agency’s revenue in 2023 declined by 70%, it said in the email that went out to customers, making the business model unsustainable. I fear that this is the first of many announcements as content agencies as we know them today slowly but surely get replaced by the Koalas and ZimmWriters of this weird new world.

I, for one, welcome our new AI overlords.

While I like the convenience of a tool that prompts GPT for me, I personally have a preference for the original experience of using ChatGPT (or the "Chat" mode of OpenAI's APIs).

It’s cheaper, gives me more control over the outputs, and allows me and my team to iterate and improve on our content creation workflows by constantly coming up with and trying out new prompts.

In today’s content drop, I’m going to share three of those prompts with you.

I used to be very secretive about these things until I realized that (a), we’re all in a race to the bottom—optimistic, I know—and (b), the website-building business model will likely adapt just like it has been doing since the advent of the web.

The Ghostwriter

Be my ghostwriter. I will tell you what to write, and you will write it in American English using everyday language. Can you do that for me?

This is my most used prompt.

I have turned my writers into researchers and given them a green light to accelerate their writing with ChatGPT under one condition (okay, two): they source the facts, and they instruct it what to write.

If they can write it faster than it would take them to prompt ChatGPT, they do.

But if they’re tired, distracted, or experiencing writer’s block, they revert to researching and prompting. The key to producing good content with generative AI, in my opinion, is to control the output by feeding the input.

I don’t do quick flips or earn-and-burn websites in my portfolio, so I don’t want the 💩 that comes out from "Give me a 1,000-word blog post about how to season cast iron pans" on them.

I want a well-written, fact-checked, source-enriched piece of content that cites experts (ideally, that they’ve sourced themselves from HARO or Qwoted), links out to government or academic sources, and reads like an article one would actually want to read or at least skim through.

This prompt allows me and my team to feed ChatGPT with instructions, whether in one-two sentences or in the form of bullets, on what exactly to write—and triggers it to do so in everyday language that doesn’t trigger AI detection alarms (as much) or bore the reader to death.

The output becomes a first draft, which requires editing but not necessarily fact-checking since any figures or statements are provided by us to the tool. Just like doing research before writing on a topic yourself, this helps me and my team produce better content, faster.

The Editor

Be my editor. I will give you my writing. Without changing the meaning of my text, edit it for American English and rephrase it in everyday language. Can you do that for me?

This prompt turns ChatGPT into an AI editor on steroids.

First, it's great for non-American authors—including native English speakers from Commonwealth countries—as it helps align their writing with the language of my website's main audience, readers from the United States.

Secondly, it prompts ChatGPT to simplify the text and rephrase it using plain, everyday language. This is important because a website that depends on organic search for traffic has two audiences that it needs to write for: readers and search engine bots.

Search engine bots, being bots, prefer simple, uncomplicated writing. Asking ChatGPT to rephrase it, and observing if it does, is a great way to sanity-check if you yourself are being clear. If GPT is struggling to understand your trail of thought, chances are so will Google Bot.

Sometimes, you have to dig deeper and be even more explicit in your instructions to the tool.

For instance, I have an automotive site.

This one time, I was using ChatGPT as an editor, but it was giving me way too generic suggestions—even with this prompt.

So I told it: This is way too generic. I need the text to speak to motor heads. Rewrite accordingly and remember this for all future texts I give you for editing in this chat.

Bingo.

The output becomes a second draft that needs a final round of human editing.

Notice that nowhere am I publishing non-edited, non-fact-checked content generated solely by AI. This is a choice I made since I played with generative AI in the days of Jasper, WriteSonic, and Rytr, and that I don't feel like I need to revise.

The Intern

Here’s my blog post. Imitate my style of writing and write an {introduction} / {conclusion} / {transition between these two paragraphs or sections} for it. Keep it simple, engaging, and in everyday language.

Sometimes, in your writing, you just get stuck.

This prompt, as long as you’re as explicit as can be in what you want out of the tool, can help you get unstuck, especially for the "fluffy" parts of an article, be it the intro, the conclusion, or a transition between paragraphs or sections.

P.S. If you're not a LowFruits user *yet*, now is a great time to become one—you can get started for free. Two of my case study updates are about to go out on the LowFruits blog this month, and LowFruits users will be the first to hear about them (on email).